This change makes a bunch of changes to modernize our JavaScript build

process across our products:

Better handle npm workspaces.
Prior to this change, we had a reviewboard/package.json file which was

the "real" JS package file, and a symlink from the repository root into

that (for the purposes of npm install). This then also specified a

"workspaces" key which had .npm-workspaces/*, in which we'd

automatically create symlinks for djblets and reviewboard. This ended up

kind of double-installing packages, where we'd have both the top-level

reviewboard as well as a reviewboard inside the workspaces.
This change removes the symlink, and creates a new top-level

package.json file which just specifies the workspaces and scripts. All

actual dependencies now live only inside the reviewboard/package.json

file.
This allows creating additional symbolic links in the workspace

directory in order to bring in development copies of spina, ink,

beanbag-eslint-plugin, or whatever other JS dependencies on desires.
Depending on which JS packages one wants to include in their workspace,

this may require having a shared parent directory and a symbolic link

from ../node_modules to ./node_modules, since some things like to

resolve to absolute paths and then crawl up the directory tree (looking

at you, tsc).

Update and consolidate JS dependencies.
Several updates have been made to our JS dependencies:

We now rely on the new @beanbag/js-buildkit metapackage for our

  eslint and typescript dependencies.
babel-polyfill has been replaced with core-js and the

useBuiltIns setting for @babel/preset-env (as per Babel's

  documentation). This ends up transforming the core-js library to only

  include what's needed for the browserslist target.
Most other dependencies have been updated to the latest. CodeMirror

  is updated within its major version, since version 6 is completely

  different.
Update to Storybook 10.
Update to ESLint 9.


Stop vendoring dependencies.
We had a number of things that were vendored into our source tree, many

from before we used node_modules at all.

We were both getting jasmine-core through npm as well as vendoring a

  copy of jasmine 5.1.0 into our lib/js directory. I've changed it so the

  js-test-libs index pulls in the relevant jasmine boot files from the

  installed jasmine-core, making it easier and more reliable to upgrade in

  the future. We also had vendored in sourcemapped-stacktrace and

  masonry-layout, which I've changed to also come entirely from npm.

I've also ported our old jasmine-hide-filtered and jasmine-sourcemaps

  code to TypeScript.


jquery-ui's JavaScript was being pulled from node_modules, but we

  had vendored in a copy of the CSS and some of the images. Fixing this

  required some build-time shenanigans inside settings.py to find the

  relevant files.


Masonry (used in the admin dashboard) is a very simple update,

  especially since there haven't been any new versions at all.



Simplify browserslist.
Instead of a complicated browserslist config, I've changed it to just

specify baseline widely available. This ends up pulling in a few

more older versions of browsers than our old config did, but it seems

like it's the right choice in the current age.

Fix an issue in the page-banner CSS.
The newer version of LESS correctly identified :not('.ink-c-button')

as an error. I've removed the quotes so that CSS builds properly.

Smoke tested the dev server.
Built static media and verified that everything looked correct,

  especially the core-js polyfills and jQuery-UI stylesheets.
Ran python tests.
Ran js-tests.
Ran eslint (eek).
Built the rbintegrations extension to verify that static media

  building still worked for that.
Ran build-storybook
Ran the storybook and flipped through the couple components that are

  there. Verified that things worked as expected.
Verified that things that used jQuery-UI (comment dialog, autocomplete

  fields) looked correct.